Structure Reveals a Mechanism of CRISPR-RNA-Guided Nuclease Recruitment and Anti-CRISPR Viral Mimicry.

CONTACT FOR REAGENT AND RESOURCE SHARING

Requests for materials should be addressed to Blake Wiedenheft ( bwiedenheft@gmail.com ).

EXPERIMENTAL MODEL AND SUBJECT SHARING Microbes

Escherichia coli cells were cultured on LB medium.

METHOD DETAILS Protein expression and purification

P. aeruginosa Csy complex Csy genes and a synthetic CRISPR were co-expressed on separate vectors in E. coli BL21 (DE3) cells as previously described ( Rollins et al., 2017 ). Expression was induced with 0.5 mM isopropyl-D-1-thiogalactopyranoside (IPTG) at an optical density (OD 600nm ) ~0.5. Cells were incubated overnight at 16°C, then pelleted by centrifugation (5000 × g for 15 min at 4°C) and re-suspended in lysis buffer (50 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) pH 7.5, 300 mM potassium chloride, 5% glycerol, 1 mM Tris(2-carboxyethyl) phosphine hydrochloride (TCEP), 1× protease inhibitor cocktail (Thermo Scientific)). Pellets were sonicated on ice for 3 × 2.5 min (1 sec on, 3 sec off), then lysate was clarified by centrifugation at 22,000 × g for 30 min at 4°C. The Csy complex self-assembles in vivo and the intact complex (with N-terminal 6-histidine affinity tags on Cas7f) was affinity purified over NiNTA resin (Qiagen) which was washed once with lysis buffer supplemented with 20 mM imidazole before elution with lysis buffer supplemented with 300 mM imidazole. Protein was then concentrated (Corning Spin-X concentrators) at 4°C before further purification over a Superdex 200 size-exclusion column (GE Healthcare) in 20 mM HEPES pH 7.5, 100 mM KCl, 5% glycerol, 1 mM TCEP. P. aeruginosa Cas1–2/3 complex The Cas1–2/3 complex was expressed and purified using previously described methods and the plasmids are available on Addgene (#89240) ( Rollins et al., 2017 ). Briefly, the expression vector was transformed into E. coli BL21 (DE3) cells, and the cells were induced with IPTG at an OD 60 o of 0.5. Expression was induced with 0.5 mM IPTG at OD 600 = 0.5 nm. Cells were pelleted and lysed as described above. Co-expressed Cas1 (with N-terminal 6-histidine affinity tag) and Cas2/3 (untagged) were affinity purified using NiNTA resin (Qiagen), which was washed once with lysis buffer supplemented with 20 mM imidazole before elution with lysis buffer supplemented with 300 mM imidazole. Protein was concentrated (Corning Spin-X concentrators) at 4^3 before further purification over a Superdex 200 size-exclusion column (GE Healthcare) in 20 mM HEPES pH 7.5, 100 mM KCl, 5% glycerol.

Electron microscopy Grid preparation for cryo-electron microscopy

Prior cryo-EM studies with the Csy-Acr complex ( Chowdhury et al., 2017 ) showed that Csy complexes adopt a preferred orientation in ice. Addition of 0.05% (v/v) Lauryl Maltose Neopentyl Glycol (LMNG, Anatrace) to the sample helped in overcoming this orientation bias problem. 4μL of 2 mg/mL purified Csy-DNA complex, mixed with 0.05% (v/v) LMNG was added onto freshly plasma cleaned (hydrogen, oxygen plasma) 300 mesh UltrAuFoil R1.2/1.3 holey Gold grid (Quantifoil). After manually blotting off excess sample with a Whatman No.1 filter paper for 5–7 s, the sample was immediately vitrified by plunge freezing in liquid-ethane at −179°C. The entire cryo grid preparation process was carried out at 4°C and 98% relative humidity to minimize excessive evaporation of sample from grid surface.

Cryo-electron microscopy data acquisition

Cryo grids were loaded into a 200keV Talos Arctica (Thermo Fisher) transmission electron microscope. 3,208 micrographs ( Figure S1A ) were acquired with a K2 Summit (Gatan) direct electron detector operating in super-resolution mode, using the Leginon automated data collection software ( Suloway et al., 2005 ) at a nominal magnification of 36,000X (super-resolution pixel size of 0.575 Å/pixel; physical pixel size of 1.15 Å/pixel). Each micrograph was collected as dose-fractionated movie, where each movie comprised of 56 frames acquired over 14 s with a cumulative exposure of ~58 electrons/Å 2 . A nominal defocus range of 0.6mm to 1.5mm was used for collecting the data.

Image processing and 3D reconstruction

The super-resolution movie frames were first Fourier-binned 2 × 2 times to a pixel size of 1.15 Å/pixel, prior to dose-weighted frame alignment using MotionCor2 ( Zheng et al., 2017 ) implemented in the Appion ( Lander et al., 2009 ) image processing workflow. CTF parameters for the summed aligned micrographs were estimated using CTFFind4 ( Rohou and Grigorieff, 2015 ) ( Figure S1B ) and only micrographs with confidence values above 90% were further processed. Particles were picked from these micrographs using the FindEM (Roseman et al., 2004) template-based particle picker in the Appion workflow, using selected 2D class averages from the previous Csy-Acr complex dataset as templates ( Chowdhury et al., 2017 ). Coordinates from these picks were then imported into RELION 2.0 ( Kimanius et al., 2016 ), and 1,543,677 particles were extracted with a box size of 288 pixels, which were binned by a factor of 2 (resulting box size 144 pixels, pixel size of 2.3 A/pixel). These particles were then subjected to reference-free 2D classification ( Figure S1C ) within RELION 2.0, and a stack of 962,677 particles was obtained by selecting classes that represented different orientations and contained high-resolution features. These selected particles were subjected to 3D refinement ( Figure S2A ), using a 60 Å low passed filtered Csy-Acr map (EMD-8624) as an initial model. Particles from the 3D refinement were subjected to 3D classification without alignment and sorted into four classes. 743,861 particles belonging to two well-resolved 3D classes with the intact Cas8f C-terminal helix bundle were selected for further processing. Based on the x and y shifts associated with these particles, unbinned particles (box size 288 pixels, and pixel size of 1.15 Å/pixel) were extracted with re-centered coordinates. These particles were subjected to unmasked 3D refinement followed by another round of refinement with a soft edged 3D binary mask. The mask used for the refinement was generated using the volume from unmasked refinement run, that was expanded by 5 pixels with 8 pixels Gaussian fall-off smoothing. All subsequent masks that were used for downstream data processing were generated using the same procedure. The resulting reconstruction reported a resolution of 3.85 Å at a Fourier Shell Correlation (FSC) of 0.143. To further sort structural heterogeneity, particles from this 3D refinement were subjected to three class 3D classification without alignment. 291,227 particles from the best resolved 3D class of the full complex (containing the helix bundle of Cas8f) were further refined, resulting in a 3.4 Å resolution (at an FSC of 0.143) reconstruction ( Figure S3G ). Though the majority of this reconstruction presented well-defined structural details, the head, tail, and the helix bundle region of the Csy-DNA complex were poorly resolved due to intrinsic flexibility ( Figure S2A and S2C ). In order to improve the quality of the map for the different regions of the Csy-DNA complex we used the signal-subtracted focused classification and refinement technique ( Figure S2B ) in RELION ( Bai et al., 2015 ; Chowdhury et al., 2017 ). The whole complex was divided into three regions with some overlap between contiguous regions. These were the head-Cas8f helix bundle-Cas7f.1-Cas7f.2 subunits (region-1), the backbone comprising of all six Cas7f subunits and target DNA bound crRNA (region-2), and the tail-Cas7f.6 subunits (region-3). Each of the signal-subtracted particle stacks were subjected to independent 3D refinement and clustering (classification without alignment) runs, resulting in better quality map for each of the three regions. The final focused map for the head-Cas8f helix bundle-Cas7f.1 subunits, tail-Cas7f.6 subunits, and the backbone region were resolved to 3.3 Å, 3.2 Å and 3.1 Å (at 0.143 FSC value) ( Figure S3F ), respectively. In order to better facilitate model building of the full Csy-DNA complex, the three focused maps were aligned relative to each other, with the overlapping regions and the unsharpened non-focused reconstructed map of the full complex serving as guides and alignment references. A composite map was generated from the three focused maps by retaining the maximum valued voxel at each point, accomplished by using the “vop maximum” function in UCSF Chimera ( Goddard et al., 2007 ) ( Figure S2B ). Local resolution estimations ( Figure S1E ) were calculated using the “blocres” function in the Bsoft suite ( Heymann and Belnap, 2007 ). Atomic model building The atomic models for Cas5f, Cas8f, Cas6f and Cas7f from the Csy-Acr complex (PDB ID: 5UZ9) were used as initial template models for model building. These were individually rigid-body fitted into the reconstructed maps using the “fit map” function in UCSF Chimera ( Goddard et al., 2007 ), and residue registers and backbone geometries were adjusted in Coot ( Emsley and Cowtan, 2004 ). Models for the crRNA and DNA strands were also manually built into the map using Coot. Regions of the map, particularly flexible loop regions could not be modeled due to lack of EM density. Density for the R-loop of the target DNA was not resolved well enough to observe the bases, but position and direction of the sugar-phosphate backbone was sufficient to model (see Figure S3 ). The atomic model underwent real-space refinement with rigid body fitting and simulated annealing in PHENIX ( Afonine et al., 2012 ). The refined model was used as a seed for generating 200 models in Rosetta and the top scoring model was used for further refinement. Multiple rounds of refinement of the model was performed in PHENIX and Coot to fix the geometric and steric outliers, which were identified by MolProbity during validation. Once the major issues with the model were fixed, the final refinement iterations were carried out with secondary structure and Non-Crystallographic Symmetry (NCS) restrains. The final model was subjected to a multi-model pipeline ( Herzik et al., 2018 ), which produced five models that provided a per-residue assessment of the quality of the EM density. Residues with high Ca RMSDs (>3A) were truncated to the Cp or removed from the atomic model prior to deposition. UCSF Chimera ( Goddard et al., 2007 ) and ChimeraX ( Goddard et al., 2018 ) were used for visualization and for generating all the figures for the maps and models ( Figure S3A – E and Figure S3G ). All the maps and atomic model ( Table S2 ) were deposited into EMDataBank and Protein Data Bank with accession codes EMD-9191 and PDB ID 6NE0, respectively.

Electrophoretic Mobility Shift Assays

(EMSA) dsDNA binding assay Binding assays were performed by incubating 0, 0.001, 0.01, 0.05, 0.1, 0.5, 1, 10, 100, 1000, 10,000 nM Csy complex with 3A) were truncated to the Cp or removed from the atomic model prior to deposition. UCSF Chimera ( Goddard et al., 2007 ) and ChimeraX ( Goddard et al., 2018 ) were used for visualization and for generating all the figures for the maps and models ( Figure S3A – E and Figure S3G ). All the maps and atomic model ( Table S2 ) were deposited into EMDataBank and Protein Data Bank with accession codes EMD-9191 and PDB ID 6NE0, respectively.

Electrophoretic Mobility Shift Assays

(EMSA) dsDNA binding assay Binding assays were performed by incubating 0, 0.001, 0.01, 0.05, 0.1, 0.5, 1, 10, 100, 1000, 10,000 nM Csy complex with